ES2434015T3 - Terminal Location Technique - Google Patents

Abstract

A method of estimating the location of a mobile terminal (1) in a mobile telephone communications network having a plurality of base stations (3 - 5, 7 - 9), including the method: determining the measurement of at least one parameter signal for the mobile terminal in relation to each of the plurality of base stations; using the signal parameter measurements to obtain a first estimate of the distance of the mobile terminal from each of the plurality of base stations; and characterized by using at least one estimator in each of the first distance estimates to obtain a refined distance estimate of the mobile terminal (1) from each of the plurality of base stations i such that the at least one estimator matches at least to one of the following formats **Formula**

Description

Terminal Location Technique

This invention relates to a technique for locating and / or tracking a mobile terminal and associated equipment. More particularly, the present invention relates to the location and / or tracking of a mobile communication terminal, the terminal itself being in a mobile telephone communication network and the associated communication network equipment. More specifically, the present invention relates to a system for and a method of estimating the position of a mobile terminal in a mobile telephone communications network.

With the increasing capacity and area coverage of cellular networks worldwide, mobile phone telecommunications providers are able to provide new and better services for their subscribers. In particular, one type of service in which there has been a growing interest in the latter are Location Based Services (LBS).

The LBS provides a mobile terminal user with location-based content provision, such as local weather forecasts, and may also allow the mobile terminal user to identify persons or establishments near their proximity. For example, a mobile phone user can contact an LBS provider to identify their nearest gas station, restaurant, bank or others, based on the user's current mobile terminal (MT - Mobile Terminal) location. Similarly, finding the location of friends or family can be used, such as for security or emergency reasons, or simply meeting, by requesting an update from the location of one or more persons to the LBS provider. These services can help users, saving them time, instead of traveling through a geographical area, looking for a particular establishment / person, thus potentially avoiding traffic jams. It can also help people in emergency situations, such as helping to quickly identify their nearest hospital or the location of someone requiring emergency assistance.

These services require a location technique that can achieve reasonable position accuracy as well as commercially reasonable deployment costs. Such costs generally refer to the scope and complexity of the additional software and hardware to be incorporated into the mobile terminals and / or the network. There are also other aspects to consider, such as any penalty imposed by tracking techniques and procedures against storage / memory and battery power consumption requirements in mobile terminals.

A great variety of techniques for locating a mobile terminal have been proposed to date. The best-performing location technique in terms of accuracy and reliability is currently the Assisted Global Positioning System (A-GPS). GPS is a technique in which the terminal receives signals from a constellation of GPS satellites. From the received signals, the terminal calculates its position based on the time difference observed in the signals received from the different satellites. In the A-GPS the mobile telephone network (for example GSM) provides additional information through its air interface to assist the terminal in estimating its position (for example, differential GPS corrections). This leads to a lower “Time to First Fix” and a lower battery consumption of the terminal.

However, A-GPS is a fairly expensive approach, and the relatively low commercial acceptance of Location Based Services combined with the high costs associated with the GPS point leads to a relatively long period of time before the terminals 2G and 2G / 3G phones are equipped with GPS receivers extended. In addition, even after the introduction of GPS receivers in a large proportion of mobile terminals, since GPS requires direct contact with satellites to operate, the limited availability of satellite signals in indoor and certain outdoor environments (for example due to the effect of urban canyon) indicates the need for hybrid techniques that combine A-GPS with a location based on mobile telephony.

A simple method based on mobile telephony uses the nearest cell as a means to provide an approximate position of the terminal. For example, the cell that provides the terminal with the most powerful Received Signal Strength (RSS) measurement is determined to the terminal, and the terminal is then assumed to be within the coverage area of that cell. The accuracy of this method depends of course on the coverage area of the cell, and thus, will typically provide a mobile terminal position that is accurate in a range of 600 m to a couple of kilometers (i.e. 600 m - 2000 m depending on the cell size). This method is one of the techniques that is currently used by operators who have already introduced commercial LBS.

Terminal location techniques based on measurements of Received Signal Strength (RSS) are of reduced complexity of implementation due to the fact that RSS measurements are available on the terminal side as part of its standard operation in both standby and dedicated mode. The main problem with RSS-based location techniques is what

It refers to the precision they achieve, due to the variability of the signal, resulting from the stochastic nature of the mobile telephone radio propagation environment.

Document GB2416461 described a technique called "Location of Mobile Assisted by Statistical Terminal" or briefly "STAMP" ("Statistical Terminal Assisted Mobile Positioning"). This technique allows the use of multiple historical measurements of RSS collected while the terminal is in standby mode. During an LBS establishment phase, these RSS measurements are used with a location technique to estimate the position of the terminal, and this determination is then statistically processed to improve the accuracy of the determined position of the terminal. In this way, the "STAMP" technique was able to improve the accuracy of most existing tracking techniques, making them more useful for LBS applications. The STAMP technique, however, when applied to RSS measurements, is also susceptible to signal variability, particularly with regard to the effect of multiple routes that affects its accuracy. In GB2416461, this problem was solved by eliminating measurements with a high error content (for example, due to rapid fading).

Signal variability is particularly a problem when using STAMP with empirical models such as Hata or Cost231, since these models estimate the distance between the terminal and the base station, distance being an exponential function of RSS. Therefore, undue variability in RSS measurements will result in drastically exaggerated estimation errors.

Although it has been found that the STAMP technique provides a reasonable degree of position accuracy, and to be implementable for a reasonable cost, the position accuracy that is currently achieved could still be improved, particularly with respect to mitigation of position error. of the terminal introduced by RSS-based techniques.

The method proposed in EP135393 which is used for the delimitation of two parts refers to the ability to use a series of measurements of the Received Signal Strength (RSS) collected while the terminal is connected to the network (that is, during a call or a data session) in order to improve location accuracy. The techniques proposed in document EP135393 correspond to a filtering approach that could be considered as an alternative to the Kalman Filter. However, the method used in this document for the estimation of distance based on RSS does not consider the use of estimators that optimize the accuracy of the location of the terminal taking into account the filtering process applied.

The method proposed in WO2006096923 takes advantage of a generic route loss model (based on Hata's formulation, Cost231 of the route loss estimate) and thus, is based on the location of the terminal to provide a method for identifying the Specific route loss model parameters that apply in the proximity of the terminal location. As a result, a refined route loss model is eventually applied in order to improve the accuracy of the location of the terminal. However, the method proposed in WO2006096923 does not address the statistical behavior of the error introduced during the estimation of the distance from terminal to base station even when the accuracy of the route loss model has been improved. In addition, the present proposed application defines statistical estimators that adapt to the Kalman Filtering process and provide greater location accuracy, an approach completely devoid of the teachings of WO2006096923.

The method proposed in WO03071303 effectively corresponds to a Weighted Minimum Square method on the RSS measurements of multiple base stations in order to optimize the resulting location of the MT. The weighted factors proposed in WO03071303 are aimed at fulfilling the expected error in the distance estimate. However, WO03071303 does not provide any analysis of the expected error in estimating the distance from the terminal to the base station and does not provide any statistical estimator that could improve the accuracy of the location of the terminal.

The method proposed in “New Method for Estimation of Mobile Location Based on Signal Attenuation and Hata Model Signal Prediction”; 2005, from OROOJI, M. et al considers a set of equations to express the location of the terminal as a function of (a) the coordinates of the base station, (b) the distances between the terminal and the relevant base stations and (c) a set of angles that relate the position of the terminal with the positions of the base stations. The method is based on an approximation of the angles required in this set of equations that allows the identification of the terminal position. However, estimating the distance between the terminal and each of the base stations does not consider any statistical estimator that could improve its accuracy. In addition, the method proposed in this publication does not provide any statistical estimator that can improve the performance of a statistical filtering process such as Kalman Filtering.

There is therefore a need for an improved network-based location technique that combines adequate accuracy with low deployment costs.

The present invention seeks to overcome or alleviate at least one of these problems.

Compendium of the invention

In accordance with a first aspect, the present invention provides a method of estimating the location of a mobile terminal in a mobile telephone telecommunications network having a plurality of base stations, as defined in the appended claims.

Conveniently, the estimator may include a Kalman filtering component.

It is also preferable that the determined measurement of at least one signal parameter is a measurement of the Received Signal Strength (RSS) and an RSS based location technique, such as Global Identity Enhanced Cell (ECGI) or STAMP, is used to obtain the first estimate of distance from the location of the mobile terminal.

Advantageously, the estimators minimize the impact of the position calculation process on the precision resulting from the location technique.

The estimators are statistical estimators that can be applied to an estimate of the distance of the mobile terminal from each of the base stations and / or applied to an estimate by trilateration / triangulation of the real position of the mobile terminal, based on the estimated distances of the mobile terminal from the plurality of base stations. The estimators are preferably inferred taking into account the nature of the route loss propagation models (such as Hata and Cost231), the error introduced in the estimation of the distance from the mobile terminal to the base station as well as the error introduced in the triangulation process

Other aspects of the invention relate to a telecommunications terminal and a communications network element, such as a base station, configured to implement the aspect of the invention, particularly as set forth in the appended claims.

Brief description of the drawings

Embodiments of the invention will now be described with reference to the accompanying drawings, in which:

Figure 1 illustrates a schematic diagram of a telecommunications network in which the invention can be implemented;

Figure 2 illustrates a flow chart of the key stages in a location determination process based on RSS;

Figure 3 illustrates a graph comparing the mean quadratic error of the MT-BS distance for three different values? Rx of variance (4 db, 8 db and 12 db) for three different estimators according to different embodiments of the invention;

Figure 4 illustrates a graph comparing the performance of the RSS estimators, according to embodiments of the invention, in an RSS based location technique;

Figure 5 illustrates a graph comparing the performance of different RSS-based location techniques that use different estimators according to different embodiments of the invention;

Figure 6 illustrates a graph comparing the performance for different estimators according to the embodiments of the invention as used in the STAMP technique.

Figure 7 illustrates a graph comparing the performance of different estimators according to the embodiments of the invention as used in the MT Tracking technique; Y

Figure 8 illustrates the comparative performance of CGI, CGI + TA and ECGI, STAMP and MT Tracking using estimators calculated in accordance with an embodiment of the invention;

Figure 9 refers to the embodiments of the invention in which Kalman filtering is applied to the different

estimators, and illustrates a graph comparing the gain factor of the Kalman filter

 based on the standard deviation of the RSS? Rx for a theoretical performance of Kalman Filtering of steady state as well as for a real simulation;

Figure 10 illustrates an Equivalent Mobile Terminal Position Calculation Model by RSS Kalman Filtering (Option A) according to an embodiment of the invention;

Figure 11 illustrates an Equivalent Mobile Terminal Position Calculation Model by Kalman Filtration of Distance MT-BS (Option B) according to an embodiment of the invention;

Figure 12 illustrates an Equivalent Mobile Terminal Position Calculation Model by Kalman Filtering of the MT Coordinates (Option C) according to an embodiment of the invention; Y

Figure 13 illustrates a graph of a comparative performance, of a simulation of the best performance estimators for the Kalman Filtered embodiments of the invention and the Basic RSS embodiments of the invention.

Detailed description of the embodiments of the invention

The key elements of a mobile phone telecommunications network and its operation will now be briefly described with reference to Figure 1.

Each cell corresponds to a respective base station (BS - Base Station, in English) of its cellular or mobile telecommunication network that receives calls from and ends calls in any mobile terminal of that cell through wireless radio communication in one or both of the domains of switched circuits or switched packets. Such a mobile subscriber terminal or User Equipment (UE) is shown in 1. The mobile terminal can be any portable communication device such as a handheld mobile phone, a personal digital assistant (PDA - Personal Digital Assistant, or a lap computer equipped with a network data card.

In a GSM mobile phone telecommunications network, the base station subsystem (BSS) includes the base transceiver stations (BTS) and the base station controller ( BSC - Base Station Controller, in English). A BSC can control more than one BTS. BTSs and BSCs comprise the radio access network.

The present invention will be described in relation to a GSM network, although it will be apparent that the concepts of the invention can be easily applied to other provisions of the mobile telephone network, including UMTS and SAE / LTE.

Conventionally, in a GSM network, the base stations are arranged in groups and each group of base stations is controlled by a mobile telephone switching center (MSC), such as the MSC 2 for base stations 3, 4 and 5. As shown in Figure 1, the network has another MSC 6, which is controlling three other base stations plus 7, 8 and 9. In practice, the network will incorporate many more MSCs and base stations than those shown in Figure 1. Base stations 3, 4, 5, 7, 8 and 9 each have a dedicated (not shared) connection to their MSC 2 or MSC 6 - typically a connection of cable. This prevents transmission speeds from being reduced due to congestion caused by other traffic.

Each subscriber to the network is provided with a smart card or SIM which, when associated with the user's mobile terminal identifies the subscriber to the network. The SIM card is pre-programmed with a unique identification number, the "International Mobile Subscriber Identity" (IMSI), which is not visible on the card and is not generally known to the subscriber. The subscriber is issued with a publicly known number, that is, the subscriber's telephone number, through which calls to the subscriber are initiated by the callers. This number is the MSISDN.

The core network 12 includes a local location register (HLR) 10, which, for each subscriber of the network, stores the IMSI and the corresponding MSISDN together with other subscriber data, such as the current or the last known Location Area of the subscriber's mobile terminal.

When a mobile terminal is in standby mode, it selects a cell to be synchronized with it (the primary cell) based on the strength of the strongest received signal. To achieve this, the terminal periodically monitors the power of the signal received from the primary cell and a set of up to six neighboring cells. As the terminal moves, the set of monitored cells as well as the primary cell may change.

Whenever a mobile terminal (which is in idle mode) selects a new primary cell, it will detect that it belongs to a different Location Area (through the Location Area Identity that each cell is emitting) from that of its primary cell previously selected. The mobile terminal will therefore send a location area update request to the core network 12 below.

With this background in mind, implementations of the embodiments of the invention will now be described, in relation to the situation of a subscriber who is using mobile terminal 1 to initiate a Location Based Service (LBS). .

When a user wishes to use the mobile terminal to initiate an LBS, the terminal will switch from standby mode to active or dedicated mode, and the establishment phase for the LBS session will begin. There are three different cases in which RSS can be used to support an LBS service:

(a) Regular Case: In this case the terminal in the phase of establishing an LBS (that is, being in dedicated mode) measures the RSS of the primary and neighboring cells and uploads this information to the network. Since the terminal is in dedicated mode, you can also provide the Synchronization Advance (TA) parameter that is made available as soon as a channel

5 communication is being adjusted to support the session. In this case, the network can use the location technique using “Enhanced Cell Global Identity” to position the terminal (that is, take advantage of the information in the service cell, the Synchronization Advance parameter and RSS measurements to carry out the location of the terminal). Alternatively, other hybrid location techniques can be applied to take advantage of

10 any other type of measurement of the mobile telephone network or other technology (eg GPS) combined with the measured RSS and the Synchronization Advance.

(b) STAMP (Statistical Terminal Assisted Mobile Positioning Mobile Location) concept: even if terminal 1 is in idle mode, it will periodically measure the RSS of its primary and neighboring cells. Do this by measuring and storing,

15 Typically in a list with timestamps for a predetermined size, the RSS measurements of your primary cell as well as your neighboring cells. In the LBS establishment phase the terminal uploads the list of the most recent RSS measurements to the network along with a single TA measurement that was made after the terminal entered the dedicated mode. The network takes advantage of the series of RSS measurements over time in order to infer a set of historical positions

20 of the terminal to provide an estimate of the actual position of the terminal. Then, by applying the statistical filtering of the resulting terminal position estimates, the network can provide a more accurate and refined estimate of the current terminal position. In addition, a wide variety of hybrid techniques can be identified if the STAMP concept is exploited depending on the additional parameters that can be well monitored in

25 sleep mode or dedicated mode.

(c) Mobile Terminal Tracking (MT - Mobile Terminal): This concept refers to a specific category of LBS in which after the establishment of the LBS session the network continuously tracks the position of the terminal based on the regular RSS and TA measurements carried out by the terminal. In this case any ECGI technique can be applied or any technique

30 hybrid that takes advantage of the available RSS and TA measurements.

The expression “base station” should be considered to encompass both standard macro base stations that support multiple cells, as well as micro or peak cells as well as Access Points such as femto cells. The number of neighboring cells that are monitored by the mobile terminal usually varies from one to six, depending on the location and density of the cells. In an urban environment, the typical number of cells

35 visible neighbors can be six due to the high density of base stations. Since the base stations support one or more cells (the typical number is three in a cellular network) the number of neighboring cells monitored corresponds to one to six different base stations.

With reference to Figure 2, a process of calculating the location of the mobile terminal based on a single set of RSS measurements involves the following steps:

40 (a) Determination of RSS measurements (or RSS and Synchronization Advance (TA) measurements (step 21);

(b) Estimating the distance from Mobile Terminal to Base Station (MT-BS), particularly based on empirical route loss modes such as Hata or Cost231 (step 22);

45 (c) Trilateration / Triangulation of the estimated distances between the Mobile Terminal and three different adjacent Base Stations (step 23); Y

(d) Determination of terminal position coordinates (step 24).

In mobile communications, the RSS of a Base Station in a certain location is variable over time and consists of two main factors, namely:

50 i) the slow fading or local environment that refers to the loss of route and the nature of the terrain. This factor is characterized by a Log-normal behavior; Y

ii) the short-term or rapid fading that is caused by the effects of multiple routes.

Rayleigh or Rician fade are the relevant models for rapid fade, depending on whether there is a line of sight or not, respectively.

According to the operation of the standard mobile terminal, the RSS is measured periodically and averaged over a given period of time. In GSM, measurements are made every 480 ms while in dedicated mode and every 2 - 5 s while in standby mode.

For terminals that are moving, rapid fading can be smoothed by averaging over 2T = 40

5 - 80 fade periods. The resulting slow fading component expressed in decibels can be modeled as a Gaussian random variable with a standard deviation of the order of 8 db - 12 db. To eliminate the impact of slow fading the signal must be averaged for longer periods of time corresponding to a distance of hundreds of wavelengths covered by a moving terminal (i.e., on the order of 100 m - 200 m for an operation in 900 MHz) In that case the averaged signal

10 provides the so-called "middle component" corresponding to the RSS which is a direct function of the loss of route that occurs due to the distance of the MT from the BS. Radio propagation loss models (for example Hata and cost231) can then be used to estimate the distance between the MT and a BS. These models indicate that the MT-BS distance is a non-linear function of RSS, a fact that greatly affects the way in which RSS errors are propagated to the estimation of the location of the MT.

15 The following formulas illustrate the Hata and Cost231 models:

Where:

-

Rx (i) is the RSS of the BS i at the location of the mobile terminal x in dbm,

-

K (i) is the transmission power of the BS i in dbm, 20 -f (i) is the frequency of the transmitter in MHz,

-

hbs (i) is the height of the BS i in m,

-

hmt is the height of the mobile terminal in m,

-

d (i) is the distance MT-BS i in km,

-

c (hmt),? (hmt) are environment functions (for example, urban, suburban) depending on the height of the

25 mobile terminal and

-

c is a specific constant for a model that depends on the radio propagation environment.

These route loss prediction models provide RSS based on the distance of the MT from the BS.

The estimate of the distance between the MT and the BS based on the RSS is determined based on inverse calculations of

The propagation path loss prediction models:

Where:

 is the resulting estimate of the distance between the MT and the BS i,

 is the RSS measured in the MT of position x. B (i): defined herein as a parameter to simplify the analysis of the equations. 35 To illustrate the statistical behavior of the estimator provided in Eq. (2) we consider the following: in the case that the route loss prediction model is properly calibrated for a geographic region

specific then we can assume that the RSS can be modeled as a random N variable distributed Gaussian / Normal. Based on Eq. (2) it can be concluded that the estimated MT-BS distance will follow a Log-normal distribution (LN):

5 where is the variance of the Gaussian distribution of the RSS,

Y

 are the average value and the variance of the Log-normal distribution  respectively. Further,

 It is defined here as a parameter to simplify the analysis of the equations.

It should be noted that the notion Rx (i) corresponds to the value of the RSS provided by Eq. (1) when the actual MT-BS distance d (i) is used. The distribution provided in Eq (3) assumes that the error of the rest of the parameters

10 involved in Eq. (1) is negligible compared to the RSS error. In addition, the above equations apply to omni directional cell antennas. From Eq. (2) the average value E, the variance V and the Mean Squared Error (MSE) of the MT-BS distance can be estimated:

From Eq. (4), since the variance and the MSE are not equal, the estimate of the distance MT-BS provided in Eq. (2) is deviated. This estimate is therefore not likely to provide much improvement in the accuracy of the distance estimate.

Distance estimator not diverted

According to a first embodiment of the invention, at a given MT-BS distance d (i), a non-deviated distance estimator has been obtained, as provided by the following equation:

Where is the factor applied to the basic estimator to get the distance estimator MT-BS

not diverted and. Actually the average value of the distance estimate is:

25 The variance and the MSE in this case are the same (property of non-deviant estimators):

Minimum Mean Square Error (MSE) distance estimator

According to a second embodiment of the invention, the minimum MSE MT-BS distance estimator can be achieved by introducing a c2 (i) factor and minimizing the following function:

The following equation shows that the MSE is optimized for the following value of c2 (i):

For this estimator, the MSE, the average value E and the variance V of the distance MT-BS are expressed as follows:

Figure 3 provides a comparison of the mean square error of the MT-BS distance for three different? Rx values (4 db, 8 db and 12 db) for the following estimators:

10 (a) basic estimator (j = 0, Ec. (2)),

(b)

non-deviant estimator (j = 1, Ec. (5) first embodiment) and

(C)

minimum MSE estimator (j = 2, Ec. (9) second embodiment).

The index "j" has been arbitrarily assigned to assist in the identification of estimators in subsequent formulas. As can be seen in Figure 3, even if the error increases with the increasing values of the

15 standard deviation of the RSS? Rx, the non-deviant estimator and the minimum MSE estimator, in particular, provide a significant improvement in the distance estimation error especially for high? Rx values.

As can be seen in Figure 2, distance determination is a component of the RSS-based terminal location technique, with another component being trilateration / triangulation. Therefore, this is another area

20 in which the calculations of the distance estimators can be refined to reduce the error in the estimation of the position.

Triangulation Error Analysis

In this aspect, triangulation is a geometric method that uses estimated distances between the MT and three different BSs to provide the position of the MT. Trilateration is a related technique that, in addition to the

25 MT position, you can determine the speed and elevation of the MT.

Triangulation is expressed by the following equations:

Where (x, y) are the position coordinates of the mobile terminal and (xi, yi) i = 1, 2, 3 are the coordinates of the base stations (typically 3 are required for triangulation calculations). Subtracting the 3rd equation (i = 3) 30 from the 1st (i = 1) and the 2nd (i = 2) from the 1st (i = 1) we get:

From ec. (12) it is feasible to estimate the position of the terminal (x, y) as follows: Eq. (13), for simplicity, can also be written as:

Where? Z and az (i), i = 1, 2, 3, z = x, and can be directly inferred from Eq. (13) and D is provided by the following equation:

Eq. (15) indicates that D is equal to 4 times the area of the triangle defined by the three BSs. It is clear that unless all three BSs are on the same line do we have D? 0, which leads to a single solution in the triangulation problem described by Eq. (14). It should be noted that this property applies to any set of

10 distances d (i) i = 1, 2, 3 even if there is an error in your estimate. Therefore the use of this approach is applicable to location based on RSS and ECGI, where the estimated distances between the MT and the three BSs definitely contain error.

The error of the position of the MT as estimated by triangulation is provided by the following equation:

15 Where j is the index of the distance estimator (that is, j = 0 corresponds to a basic distance estimator (Eq. 2)).

Assuming a given set of distances between the MT and the three different BSs d (i) (i = 1, 2, 3), the average value E and the

Variance V of the estimated MT coordinates are the following (assuming that the random variables, i = 1, 2, 3 are independent):

And the MSE in the estimation of the terminal coordinates will be (z = x, y y j = 0, 1, 2):

Terminal Coordinator Estimator not diverted

According to a third embodiment of the invention, an estimator of the position of a mobile terminal (MT) is provided, which is based on calculations of the trilateration / triangulation process. This estimator is defined to satisfy the following condition:

Where z (z = x, y) is the position coordinates of the MT (2 dimensions) and corresponds to the estimated MT position coordinates based on estimator 3.

In order to obtain a non-deviated estimate of the position coordinates of the MT (x, y), the following non-deviated MT-BS distance estimator is introduced:

The average value E, the variance V and the MSE of the MT coordinate estimation after triangulation results:

As expected, the MSE is equal to the variance in this case due to the non-deviated nature of the estimator.

10 Terminal coordinate MSE estimator

According to a fourth embodiment of the invention, an MSE estimator is provided below the optimum of the mobile terminal position (MT), which is based on the result of the triangulation process, using an estimator of appropriate distance.

An estimator below the optimum is based on Eq. (18), where the MSE of the 15 MT coordinate estimate is reduced by minimizing the following quantities:

Based on Eq. (18), the minimization of the amounts of Eq. (21) takes place when the following estimator is applied:

20 In this case, the average value and variance of the estimated terminal coordinates will be:

And the resulting MSE will be:

The estimator in Eq. (22) is below the optimum to optimally minimize the MSE of the 25 MT coordinates; apply the following set of simultaneous equations:

However, the resulting solution for cj (i) leads to a function of the distances d (i) of real MT-BS for which the parameters are not known, so that Eq. 21 is not a practical solution.

Terminal position estimation error

To illustrate the effectiveness of the non-deviated terminal coordinate estimator, the error in the estimation of the terminal position will be analyzed based on the error introduced in the estimation of the terminal coordinates using the triangulation method using this estimator. In this aspect, the error based on the position of the terminal can be estimated as:

10 Since no obvious distribution can be inferred for the previous parameter, it is considered an approximation through the following Rice distribution:

The previous approach is based on the following assumptions: it is assumed that they are independent random variables with Gaussian distribution with average values

 respectively and common variance 15de. The mean value and variance of the MT location error can therefore be approximate as follows:

The ue (j) parameter is provided by the following equation:

20 To estimate ue (j) for a given MT location (for example distances d (i), i = 1, 2, 3 given), Ec (16) can be exploited, which defines the position error MT as estimated by triangulation. For the

that concerns the parameter, since the assumption of common variance for and

 Invalid, the following approximation is applied based on the definition of the variance of the error and the Ecs. (26), (28):

25 From Eqs. (28), (29), (30) we obtain:

Parameters

Y

for a given MT position they are provided by Eq. (17) as:

The application of the MT coordinate coordinate estimator (j = 3, see also Eq. (19)) leads to a value

zero mean both for

as for. In this case, ue (j) is also zero and the distribution of the MT location error can be approximated by a Rayleigh distribution.

To illustrate the effectiveness of the estimators, a simulation model has been developed. The model considered a geographical area with a set of base stations that each support three cells, each cell corresponding to a sector of 120 degrees of equal radius. During the simulation the position of the mobile terminal was randomly estimated from the Hata model. A white Gaussian noise with variance was then added

to the resulting RSS and the coordinate estimation process of the mobile terminal suffered, using the base estimator (j = 0), the non-deviated estimator (j = 1) and the minimum MSE estimator (j = 2).

The estimated position of the terminal that was obtained was then compared with the position of the actual terminal to determine the location error (Eq. (26)). In this aspect, Figure 4 provides the average error resulting from the estimation of the position of the mobile terminal using the different estimators presented in this document (j = 0, 1, 2, 3, 4) for a variety of? Rx (4 db, 8 db and 12 db) and for a cell radius of 1500 m. To summarize the estimators shown in Figure 4:

 The estimator 0 (j = 0) is the basic deviated distance estimator of Eq. 2;

 The estimator 1 (j = 1) is the distance estimator not diverted from Eq. 5;

 The estimator 2 (j = 2) is the distance estimator of the Minimum MSE of Eq. 9;

 The estimator 3 (j = 3) is the coordinate estimator of the non-deviated terminal of Eq. 19; Y

 The estimator 4 (j = 4) is the estimator of the MSE of terminal coordinates below the optimum of Eq.

22

Figure 4 indicates that the MSE estimator of MT coordinates below the optimum (j = 4) provides the best accuracy. Also, the minimum distance MSE estimator (j = 2) behaves better than the non-deviated MT coordinate estimator (j = 3). It is clear from Figure 4 that a significant advantage is achieved by the estimators of the present invention, specifically for high values of the standard deviation of the RSS (? Rx). However, even the best performance estimator (j = 4) leads to an accuracy of the order of 500 m.

In order to show the applicability of these estimators for RSS-based localization, a set of alternative techniques have been analyzed, namely: CGI (Global Cell Identity), CGI + TA and ECGI.

According to the CGI technique, only the Cell Identity is taken into account, so it has been assumed that the MT is located at the centroid of the cell coverage.

In the CGI + TA technique the distance between the MT and the BS is estimated based on the Synchronization Advance parameter. According to the GSM specifications, the Synchronization Advance parameter is an integer with a range from 0 to 63 and allows the MT while in dedicated mode to estimate the correct synchronization of the transmission to the BS. A Synchronization Advance unit corresponds to a distance of approximately 550 m. Therefore, for this technique we have assumed that the Synchronization Advance corresponds to the delay exclusively caused by the MT-BS distance so that the distance is approximated by the following equation:

Therefore, in the CGI + TA technique it has been assumed that the position of the MT is at a distance dTA and at an angle identical to the azimuth of the cell antenna.

In ECGI, the distance from the BS of the primary cell is also estimated based on the Synchronization Advance according to Eq. (33). Then, based on the RSS, the distance from two other BSs is estimated and triangulation is applied. Therefore, the analysis for RSS-based MT-BS distance estimators also applies to the ECGI technique.

The accuracy of these techniques, namely CGI, CGI + TA, based on RSS and ECGI is compared in Figure 5 for? Rx = 8 dB. RSS-based techniques (ie, RSS and ECGI) are provided for basic and best-performing estimators (j = 0 and j = 4 respectively). Figure 5 indicates that the RSS and ECGI techniques that use the basic deviant estimator provide less precision than the simplest CGI and CGI + TA techniques. The basic estimator is therefore not a commercially attractive approach. The best performance estimator (j = 4) significantly improves location performance based on both RSS and ECGI; however, the resulting precision is still of the order of precision provided by CGI and CGI + TA respectively. In this context, the commercial interest to apply location based on RSS or ECGI is quite questionable.

Therefore, according to another embodiment of the invention, the estimators are used in the STAMP technique described in GB2416461. In this aspect, this technique collects a series of RSS measurements while the MT is in standby mode. The process of estimating the position of the terminal according to the STAMP technique in this example is carried out by applying an adaptive Kalman filter to the estimated terminal position coordinates (ie, the process output of Figure 2) . The main assumption required by the Kalman filter in this case is that the difference between the estimated and actual coordinates of the terminal follows a normal zero distribution. As shown in the analysis of the previous section of the document, this assumption can be approximated using the MT-BS distance estimator that leads to a non-deviated estimate of the terminal coordinates (see Eq. (19)).

Figure 6 represents the performance of the STAMP technique for the five different estimators (j = 0 to 5). This graph shows that estimator 3 (that is, the coordinate estimator of the non-deviated MT) is the one that behaves best, with an average location error of the order of 200 m (see Figure 4, where the average location error for the best estimator of the location based on the RSS it was of the order of 550 m).

Figure 7 illustrates a graph showing the performance of the MT Tracking technique for each of the estimators j = 1 to 4. As described above, the MT Tracking is a technique in which after the establishment phase In an LBS session, the network constantly monitors the position of the Mobile Terminal, taking advantage of regular RSS and TA measurements. Therefore, compared to the STAMP technique in MT Tracking, RSS measurements are more frequent (every 480 ms in GSM) and the TA parameter is always available (since the terminal is in dedicated mode).

Figure 8 provides a global comparison of the different techniques, showing the overall difference between the approach that behaves worse (i.e., RSS-based location using Estimator 0) and the techniques that behave better, STAMP and MT Tracking using Estimator 3.

More specifically, Figure 8 provides a comparison between the accuracy provided by location based on RSS (Estimator 1), CGI, CGI + TA, ECGI (Estimator 4), STAMP (Estimator 0 and 3) and MT Tracking. As can be seen in this figure, the performance of the non-optimized STAMP (that is, with Estimator 0) raises the CGI method despite the fact that it provides significant location improvements based on basic RSS.

The optimized STAMP (that is, when taking advantage of the non-deviated MT coordinate estimator j = 3) clearly behaves better than the CGI, CGI + TA and ECGI techniques. In addition, the location accuracy of the resulting MT for this optimized STAMP is in the range of 100 m - 200 m even for high cell radius values (Figure 8) and also for high standard RSS deviation values (Figure 6) . The performance of the STAMP technique corresponds to an improvement of approximately 50% accuracy compared to the CGI, CGI + TA and ECGI techniques.

This allows optimized estimators used in conjunction with techniques such as STAMP to provide advantageous results, and according to this they are likely to be very promising in terms of commercial exploitation. STAMP is a particularly advantageous technique, since it can only be implemented using software modifications, which reduces deployment costs.

The accuracy of the MT Tracking based on estimator 3 (that is, using estimator 3 to estimate the distances between the MT and the second and third BSs and the Tracking Area (TA) known for the distance estimation from the first BS) is greater than the optimized STAMP for two main reasons:

(to)

the use of the TA that is our simulation model is always considered correct (something that is not always true in real networks) and

(b)

the shortest sampling period applied to MT tracking (480 ms for an active session in GSM compared to 2 - 5 s during the MT standby mode).

Above all, Figure 8 shows the empirical benefits that can be achieved by the estimators of the present invention, particularly when combined with existing position determination techniques, such as STAMP and MT tracking.

Although the analysis provided in the simulation model has used some assumptions that are not fully applicable in a real network (for example, omni directional antennas), it is considered however that the model is a good indication of the theoretical accuracy that can be achieved by location based on RSS.

The estimators of these embodiments of the present invention have been inferred, taking into account the nature of the path loss propagation models (such as Hata and Cost231) and also the error introduced in the estimation of the MT-BS distance as well as the error introduced in the triangulation process. Advantageously, the estimators have been shown, through Figures 6 to 8, to minimize the impact of the position calculation process on the resulting precision. The results indicate a very significant improvement in RSS-based location when applied to a single set of RSS measurements.

In accordance with another improvement of these embodiments of the invention, statistical filtering, ideally Kalman filtering or an equivalent recursive filter, is applied to the measurements / estimates used and inferred in position estimation techniques using estimators, in order to Improve your accuracy.

More specifically, in the following locations of the invention, the measurements / estimates in the estimation of the terminal position to which the Kalman filtering will be applied include the RSS measurements, the MT-BS distance estimation and the estimation of the trilateration

 The embodiments to be described below will infer the time series of the RSS measurements using STAMP and MT Tracking, which are the position determination techniques that behave better from the previous analysis.

In these embodiments of the invention, the following Kalman filtering function will be used, which assumes that the true state X in time k has evolved from the state in (k-1):

Where: ? it is the state transition model applied to the previous state Xk-1; ? it is the control input model that is applied to the control vector Wk; Yk is a measurement (or observation) in time k of the true Xk state; M is the observation model that maps the true state space into the observed space; and Uk is a random vector that represents the measurement noise.

The vectors Xk and Yk can be represented as:

where N is equal to the number of base stations for which RSS measurements have been inferred.

The parameters defined in the vectors of Eq. (35) differ depending on the measurement / estimate to which the Kalman filter has been applied. For example, to:

(a) Kalman filtering of the RSS i = 1, 2, 3: xk (i) = Rx (i), the RSS of the BS i, v k (i) is the speed of

change of the RSS of the BS i (that is, the rate of change of Rx (i)), and is the measured RSS; (b) Kalman Filtering of Distance MT-BS i = 1, 2, 3: x k (i) = d (i) the distance between the MT and the BS i,

vk (i) is the rate of change of d (i), and is the estimated MT-BS distance based on the measured RSS; Y

(c) Kalman filtering of the MT Coordinates i = 1, 2: xk (1) = x, xk (2) = y, the 5-position coordinates of the MT, v k (1) = vx (tk), vk (2) = vy (tk) is the speed of the horizontal and vertical MT and

, are the MT coordinates estimated based on the measured RSS.

Further, ?, ? and M are matrices of (2Nx2N), (2NxN) and (Nx2N), respectively, which can be represented as follows:

10 Where IN is the unit matrix of NxN and 0N is the matrix of zero NxN and? T corresponds to the sampling period of the measurements. N is the number of measurements to be analyzed. Typically, in the Kalman filtering of the RSS N = 3, in the Kalman filtering of the Distance MT-BS N = 3 and in the Kalman filtering of the MT Coordinates N = 2.

In addition, Wk is a middle zero Gaussian distributed random vector parameter that represents the process noise with a covariance matrix Q:

where (i = 1,…, N) represents the variance of the rate of change of the RSS of the BS (i) (that is, v k (i)):

And finally, Uk is also a mean zero Gaussian random vector parameter that represents the measurement noise with a covariance matrix R:

20 Where is the variance of the RSS measurements (i = 1, .., N).

This Kalman Filtering function will now be used together with the different estimators described in the previous embodiments of the invention to illustrate how Kalman Filtering can be used to further optimize the accuracy of the location of the MT.

It will be clear, however, that the previous analysis that led to the estimators as described in the

25 Equations 5, 9, 19 and 22 cannot be validly applied directly to Kalman filtering. Therefore, properly reviewed estimators that optimize the accuracy of the MT position in the presence of Kalman filtering need to be diverted.

In this regard, one of the key elements for understanding the role of Kalman filtering in the MT position estimation process is the performance of the filtering process under stationary conditions. With the final purpose

30 to have a convergent steady state behavior the following conditions must be met:

(a) The pair {?, M} in the state space of the Kalman filter of Eq. (34) must be fully observable:

This condition can be shown as valid based on the fact that:

(b) The pair {?, G1 / 2} must be fully controllable:

And therefore:

10 It can be proved that Eq. (41) is fulfilled based on Eq. (43).

Since both Eq. (39) and Eq. (43) are met for the state space of the Kalman filter of Eq. (34), then there is a single positive solution P in the following Riccati algebraic equation:

The solution P of this equation of Riccati algebraic determines the covariance matrix:

And Kalman's K gain from the Stationary State will be:

It will be clear that the analytical solution of Riccati's Eq. (44) requires the estimation of the Eigen values of the following matrix:

Where: In the analysis and simulation carried out in this report, an approximation of Riccati's Eq. (44) was obtained using MatlabTM analysis software.

In addition, defining the revised estimators the following set of assumptions regarding the impact of Kalman Filtering (KF - Kalman Filtering) has been adopted:

(a) In terms of statistical behavior, the pdf (probability density function - Probability Density Function, in English) of the output of Kalman Filtering can be approximated by the pdf of the entered parameter;

(b) in the steady state the average value of the KF output will be equal to the average value of the 10 KF input:

(c) in the steady state the variance of the KF output will be less than the variance of the KF input, which represents the gain of the K filter:

15 Based on the equations of the KF state space (34), the steady state variance of the KF output of Eq. (51) can be rewritten as follows:

Where 0 <ƒ (i) <1 represents the gain of the Kalman K Filter of the steady state for the parameter xk (i) (i = 1,…, N).

Such Kalman Filtering will now be used in three different embodiments of the invention, namely, with reference to Figure 2, being applied to the output of the RSS 21 measurements (Option A), to the Distance Estimation MT- BS 22 (Option B) and the Estimation of the Position Coordinates of the MT after Trilateration 23 (Option C) and the appropriate combined estimators defined.

Option A: Kalman's RSS filter

25 The measured RSS is considered to follow a Normal distribution:

This condition maintains the assumption that a It applies to all Base Stations involved in the location of the MT. Based on this assumption it can be concluded that in the steady state the series resulting from the filtered RSS measurements will follow a Normal distribution with the same variance and lower:

Figure 9 comprises the gain factor of the Kalman Filter

 versus? Rx based on the KF performance of the theoretical steady state, as well as according to the simulation that will be explained below. This figure shows that the simulation closely follows the theoretical performance, decreasing the KF gain of the RSS exponentially as the standard deviation of the RSS measurements increases.

5 Based on the above analysis, it is feasible to consider a MT mobile calculation process that incorporates the Kalman Filtering of the RSS (option A - see Figure 10). As can be seen in Figure 10, in the simplified equivalent process model the impact of the Kalman Filter has been replaced with the appropriate statistical characteristics of the equivalent RSS measurements. The validity of this equivalent model will be illustrated in the simulation that will be outlined below.

10 This model in Figure 10 can be used to produce the revised MT-BS distance estimators that conform to the Kalman Filtering of the RSS (option A). In this regard, the MT-BS distance estimators reviewed in the case of Kalman's RSS filtering will be the following:

(to)

Estimator 0: Basic estimator for the distance MT-BS as provided by Eq. (2).

(b)

Estimator 1: The MT-BS distance estimator Not diverted:

(C)

Estimator 2: The MSE estimator of the minimum MT-BS distance:

(d)

Estimator 3: The MT coordinate estimator Not diverted:

(and)

Estimator 4: The MSE estimator of the MT coordinates below the optimum:

It should be noted that a KF gain of corresponding to the case in which Kalman Filtering has not been applied, and the estimators defined above are equal to the estimators defined for the case of location of the basic RSS MT (as defined in Equations 5, 9, 19 and 22).

25 Option B: Kalman Distance Filter MT-BS

This option incorporates filtering after estimation of the MT-BS distance, as illustrated in Figure 11. Figure 11 also shows a model for calculating the position of the equivalent simplified MT in which the Kalman Filter has been replaced with the appropriate statistical characteristics of the MT-BS distances. The validity of this equivalent model will be illustrated in the simulation that will be outlined below.

30 This model in Figure 11 can be used to produce the revised distance estimators that incorporate Kalman Filtering (option B). In this aspect, the MT-BS distance is considered to follow a Lognormal distribution:

The impact of Kalman filtering on the probability density function of the filter output can be approximated by the following:

From the same analysis in the steady state of the Kalman Filter we should have:

Where fd (i) represents the gain of the Kalman filter. From Eq. (61) we have:

And from Eq. (62):

Based on Eq. (63), Eq. (64) results:

And based on Eq. (65) we get from Eq. (63):

This model can now be used to produce the MT-BS distance estimators that correspond to option B of the MT-BS Distance KF.

15 Based on these results, it is possible to define the revised MT-BS distance estimators that are appropriate for this MT-BS Kalman filtering method.

(to)

Estimator 0: Basic estimator for the distance MT-BS: as provided by Eq. (2).

(b)

Estimator 1: The MT-BS distance estimator is defined as follows:

20 Therefore, the MT-BS distance estimator is identical to that applied in the event that no Kalman Filtering (Eq. (5)) is applied due to the fact that Kalman Filtering does not affect the average value of agreement with the assumption of the previous section (Eq. (50)).

(c) Estimator 2: the MSE estimator of the distance MT-BS Minimum: defined as the one that minimizes the following:

From Eq. (64), the minimum MSE is achieved by the following estimator:

So estimator 2 is:

(d) Estimator 3: The estimator of the coordinates of the MT Not diverted: From the analysis of the process of trilateration, to get the estimate of the coordinates of the MT not diverted, the following condition must be applied:

10 So estimator 3 is:

(e) Estimator 4: The MSE estimator of the MT coordinates below the optimum: according to the analysis of the trilateration process this estimator should minimize the following amount:

15 Based on the assumption that:

the minimization of the MSE (Eq. (72)) is achieved for: So estimator 4 is:

It should be noted that for a KF gain of ƒd (i) = 1 (corresponding to the case that Kalman Filtering

5 does not apply to everything and therefore no profit is produced) the previous estimators are identical to those applicable for the non-filtering process (location based on basic RSS), as defined in Equations 5, 9, 19 and 22

Option C: Kalman filter of the MT coordinates

This option incorporates filtering after the trilateration stage, as illustrated in Figure 12. Figure 12 also

10 shows an equivalent simplified MT position calculation model in which the Kalman Filter has been replaced with the equivalent MT-BS distance statistical characteristics. The validity of this equivalent model will be illustrated in the simulation that will be outlined below.

This model in Figure 12 can be used to produce the revised estimators that correspond to Kalman Filtering as for option C. In this respect, the coordinates of the MT (x, y) after the stage of

15 trilateration can be expressed as follows:

Where for (xi, yi) i = 1, 2, 3 the coordinates of the Base Stations considered for the trilateration process are:

20 The probability density function of the MT coordinates is not known, since it corresponds to the sum of the Log-normal distributed parameters. Therefore, in this option of Kalman Filtering we adopt the model of calculation of the position of the equivalent MT represented in Figure 12. As can be seen in this Figure it is assumed that the model of the calculation of the position of the equivalent MT is based on a set of equivalent MT-BS distances that follow a Log-normal distribution with appropriate mean and variance parameters. For

In inferring the characteristics of the equivalent MT-BS distances, the following conditions are used, which are based on the assumptions presented in Equations (26) and (27) as regards the steady state of the Kalman Filter:

In particular Eq. (79) leads to the following condition: And Eq. (80) leads to the following:

A solution that justifies both conditions provided by Ecs. (81) and (82) is as follows:

Assuming that follow a Log-Normal distribution we have, based on Eq. (84):

Thus:

10 In this case Eq. (83) leads to the following:

The equivalent MT-BS distances are therefore a function of the base station i (i = 1, 2, 3) as well as the z coordinate of the Mobile Terminal (z = x, y). This leads to different MT-BS distance estimators that can be used to estimate each of the MT coordinates according to this embodiment of

The invention.

Therefore, based on the above analysis, the MT-BS distance estimators reviewed for this embodiment of the invention are:

(a) Estimator 0: basic estimator: that provided by Eq. (2).

(b) Estimator 1: MT-BS distance estimator not diverted: defined as the one that satisfies the following condition:

Based on Eq. (86) and (87) we have:

So estimator 1 is: (c) Estimator 2: MSE estimator of the minimum MT-BS distance: defined as the one that minimizes the following condition:

The above condition is minimized by the following:

So estimator 2 is:

(d) Estimator 3: MT coordinate estimator Not deviated: defined as satisfying the following condition:

A solution for the previous condition is as follows:

And based on Ecs. (86) and (87) the following estimator is obtained:

The above equation for estimating the coordinates of the MT Not diverted filtered is identical to that applicable for the location based on the basic RSS defined in Eq. (9).

(e) Estimator 4: MSE estimator of MT coordinates below the optimum: defined as the one that minimizes the following quantities:

Minimization of the above amounts can be achieved with the following:

So estimator 4 is: It is interesting to note that for a gain of the Kalman Filter of ƒz = 1 corresponding to the case that Kalman Filtering is not applied, the estimators defined for Option C are equal to the estimators defined for the case of basic RSS as defined in Equations 5, 9, 19 and 22.

5 Performance Analysis

To analyze the performance of the different estimators of the MT-BS distance filtered by Kalman Filter (that is, from Options A, B and C), a simulation model will be illustrated. This simulation model is based on a mobile telephone network with a geographical area that has a set of base stations that each support three cells of the same radius (each cell corresponds to a 120º sector of the geographical area). During the

10 simulation, the position of the MT is randomly selected within this area and the RSS levels of each of the base stations are estimated from the Hata model. Also, a Gaussian white noise with variance? Rx2 is added to the resulting RSS and a set of RSS measurements is produced in a period? T.

With respect to the mobility of the MT, a street network with the Manhattan mesh is considered and the MT is moving at a speed that is uniformly selected in a range of between 10 km / hr - 60 km / hr (typical 15 for an urban environment). At each intersection of streets the direction of movement of the MT is selected with equal probability from the possible alternative street segments, while the speed of the MT is selected again from the same interval mentioned above. Based on the time series produced from the RSS measurements, the location of the MT based on the alternative Kalman Filtering options is applied, and the position of the MT was then compared with the position of the actual MT to determine the error of location.

20 In this regard, Tables 1, 2 and 3 provide the MT location error for the three different Kalman Filtering options versus the standard deviation of the measured RSS (? Rx). This is provided for both the original estimators (i.e., the estimators 0-4, without KF) and the estimators of the distance MT-BS (i.e., the estimators 1A-4A, 1B-4B and 1C-4C, with KF ).

TABLE 1: The performance of the Kalman filtering of the RSS versus the MT-BS distance estimator applied.

Table 2: The performance of the Kalman filtering of the MT-BS Distance versus the MT-BS distance estimator applied.

5 Table 3: The Kalman filtering performance of the MT Coordinates versus the MT-BS distance estimator applied.

These tables indicate that the MT-BS distance estimators with KF provide the means to improve the accuracy of the location of the MT in all KF options. In particular, the revised MT-BS distance estimators (estimators 4A, 4B and 4C) lead to optimum performance.

10 It is interesting to note that in the case of Option B of the KF of the MT-BS distance, the performance of the estimator 1 which is identical to the estimator 1B (MT-BS distance estimator Not deviated) is very close to the performance estimator optimal (revised 4B estimator).

A similar observation applies to Option C of the KF of the MT Coordinates with respect to the performance of the estimator 3 which is identical to the estimator 3C (estimator Not deviated from the MT Coordinates). The reason for

15 This result is the way in which Kalman Filtering affects the statistical characteristics of its input (see Eq. (50) and (51)): Kalman Filtering output maintains the average value of the input and reduces the variance of the input. In the event that the KF input is not diverted then the resulting location error is equivalent to the case of a smaller? Rx. In this case the performance of all estimators is convergent, as can be seen in Tables 1, 2 and 3 for, for example,? Rx = 4 db.

20 On the other hand, in the event that the input of the KF is deviated then the resulting output maintains the deviation of the input, possibly leading to a greater error, although it has a smaller variance (due to the gain of the KF). The best performance estimator (revised 4A estimator) for all KF options provides

made a solution below the optimum to the MSE of the coordinates of the MT balancing the resulting deviation and variance.

Figure 13 compares the performance of the KF options based on their best performance estimator (4A, 4B and 4C respectively) as well as the performance of the basic RSS location (for estimator 4). As can be seen in Kalman filtering, option A remains the best performing Kalman filtering option. However, as can be seen in Figure 13, the location accuracy of the resulting MT for all KF options is in a range that is valid and acceptable for commercial use (i.e. location error of 100 m - 200 m for? Rx = 12 db).

The above analysis provides the means to estimate these parameters based on the characteristics of the steady state KF. In a commercial implementation such parameters can be calculated without connection to the Internet in order to reduce the workload of real-time processing (particularly with regard to providing a good estimate of KF gain factors).

An alternative simplified approach, which is applicable to Option B of the KF of the distance MT-BS and to Option C of the KF d of the coordinates of the MT, is to use the estimators 1B and 3C (Eq. (68) and (94) respectively). ), since they do not require the estimation of the KF gain parameters and yet, according to the previous results, show a performance very close to the best performance estimators.

With respect to applying more than one option of the KF, although such an approach is possible, it has not been found to lead to another improvement in the accuracy of the location of the MT. The same applies in the case that the revised MT-BS distance estimators are used in addition to the revised KF estimators. This result is due to the fact that, where the output of one KF is used as an input to another KF then the second KF will not provide the expected gain because its input does not show one of the characteristics that allow the KF to provide gain. That is, the series of input measurements should have measurement noise that is independent in each sample. This is a property that the output of the first KF does not have.

Therefore, in summary, the above analysis has illustrated the RSS based on the optimization of the location accuracy in the presence of Kalman filtering and has adapted the analysis of the optimization of the MT location error using basic RSS to define estimators Revised for the following Kalman Filtering options: Kalman filtering of the RSS, Kalman filtering of the MT-BS distance and Kalman filtering of the MT coordinates. After evaluating the impact of Kalman filtering on the accuracy of the location of the MT based on the steady state analysis, a set of MT-BS estimators was obtained for each of the Kalman Filtering options.

It will be clear that the exact implementation of the Kalman filter as described herein should only be considered as illustrative, and that alternative Kalman configurations are within the scope of the invention. In addition, although the use of a Kalman filter is a preferred implementation, alterative filters are possible, particularly other statistical recursive filters capable of estimating the state of a dynamic system. Examples of possible alternatives include Exponential Smoothing Filters and Dynamic Data Reconciliation Filters.

In addition, the Kalman filtering embodiments of the invention were illustrated only in relation to the RSS measurements derived from STAMP and the MT Tracking position determination techniques; however, the embodiments are applicable to any other tracking technique in which RSS measurements are collected, such as CGI, CGI + TA and ECGI.

Additionally, the embodiments of the invention have been described in relation to the monitoring of the mobile terminal and the determination of RSS measurements. This is not essential for the invention, and the same functionality can be implemented external to the terminal. For example, the terminal can send the RSS measurements to the core network for processing.

The position estimation techniques listed above are applicable to various network configurations that include 2G, 3G and 4G networks.

In addition, although the present invention has a particular applicability for RSS measurements, any other parameter suitable for communications between the mobile terminal and its adjacent base stations may be used (for example, Observed Time Difference, in English) as well as measurements associated with other technologies such as GPS. More particularly, it will be apparent that the estimators according to the embodiments of the invention are particularly applicable to the RSS parameter due to its statistical nature, and to other parameters with a similar error distribution.

Claims (9)

1. A method of estimating the location of a mobile terminal (1) in a mobile telephone communications network having a plurality of base stations (3-5, 7-9), including the method: determining the measurement of at least one signal parameter for the mobile terminal in relation to each of the plurality of base stations; use the measurements of the signal parameter to obtain a first estimate of the distance of the mobile terminal from each of the plurality of base stations; and characterized by use at least one estimator in each of the first distance estimates to obtain a refined distance estimate of the mobile terminal (1) from each of the plurality of base stations i in such a way that the at least one estimator is adapted to the less to one of the following formats: use the aforementioned plurality of refined distance estimates to obtain an estimate of the position of the mobile terminal, and apply a Kalman filter to one of the signal parameter measurements, the refined distance estimates 15 or the mobile terminal position estimate determined, where: 1..4 is the estimate of the refined distance in kilometers ? Rx2 is the variance of the Gaussian distribution of the Signal Strength Received (RSS - Received Signal Strength), , 20 is the height of the base station i in m, A (i) is an estimate of an amount of propagation path loss from the base station i; Y is the Received Signal Strength (RSS) from the base station and in the actual location of the terminal. 2. The method of claim 1, wherein the Kalman filtrate is applied by adapting the at least 25 estimator so that it fits at least one of the following formats: where:  and fz are the gain factors of the Kalman filter; Y  It is a basic estimator,. The method of any one of the preceding claims, wherein the measurement of the at least one determined signal parameter is a measurement of the Received Signal Power, RSS, and an RSS based location technique is used to get the first estimate of the distance from the location of the mobile terminal.  4. The method of any of the preceding claims, wherein the measurement of the at least one determined signal parameter is a measurement of Received Signal Power, RSS, and RSS measurements. 10 are determined from historical positions of the terminal collected when the terminal was in idle mode.

5.

The method of any one of claims 1 to 3, wherein the measurement of the at least one determined signal parameter also includes Synchronization Advance measurements, which are determined by the mobile terminal while the terminal is active.

6.

The method of any one of the preceding claims, wherein the amount of propagation path loss is estimated using a Hata model or a Cost231 model.

7.

The method of any one of the preceding claims, which also includes using the distance estimates in the triangulation / trilateration calculations to estimate the location of the mobile terminal.

A mobile terminal configured to execute a method according to any one of claims 1 to 7 of the method. 9. A communications network element configured to execute a method according to any one of claims 1 to 7 of the method. 10. A mobile telephone telecommunications network configured to execute a method according to any one of claims 1 to 7 of the method.